format() has full support for sprintf format but is written completely in D.
It is currently internal, since it is not typesafe and uses GC at one place.
After some work the function can be made public.
This commit adds tanya.range.array module which contains functions to make
the arrays act as ranges. These functions don't do any auto-decoding for
strings.
- Rename memory.op.zero to fill, which accepts one template parameter: one
byte to fill the memory with.
- Fix bug on x86_64: RAX (register keeping the value to fill with) isn't set if
the pointer was already aligned.
tanya.memory.arch.x86_64:
Linux passes the array length and the data pointer in separate registers.
Windows passes a pointer to the whole array instead (pointer to the
array length practically).
Size parameter for "make" was renamed into n, but the function
description wasn't fixed:
size => n
This commit also removes some redundant variables in "make".
Make allocators pure.
* Methods allocating/deallocating memory are pure.
* Allocator.instance is pure (once initialized, it always returns
the same instance).
* defaultAllocator getter property is pure (should be set at the
beginning, and always return the same instance after that).
Because const is transitive, if we create a range as Range!(const E)
there is no way to get the original type from inside of the range. So if
E is int*, the original type of const(E) could be const(int)* or int*.
Unqual!(const(int*)) returns const(int)*. So pass the whole container as
template parameter. It is a breaking change but since we have Range and
ConstRange aliases now, the usage should be fine.
